Our ability to predict conformation of proteins from amino acid sequences depends strongly on understanding of how proteins fold. Statistical mechanical theory of proteins has now progressed to the point that such understanding is within reach. We propose to develop a theory of protein- folding kinetic based on the idea of combining the sequence design and folding simulations within the framework of the one force-field. This allows us to disentangle the two key questions of protein science: * What is the mechanism of protein folding kinetics? * How to find the correct potential of mean force for protein folding? Our preliminary studies showed that this is a realistic approach which allowed us to address the following questions. 1. Study the nucleation mechanism of folding, i.e. determine the location and size of folding nuclei for different proteins (ubiquitin, bamase, villin). Predict and test experimentally point mutations in nucleus sites which have the most pronounced kinetic implications. Learn how to predict folding nucleus from sequence. 2. Study the pathway of folding which, after nucleus is formed, directs folding process to the native conformation as well as the distribution and structure of off-pathway "traps" which give rise to bioexponential folding kinetics. Compare predicted fast kinetic phases with experimentally observed rates and amplitudes. 3. Include side-chains packing into models, design native conformation with tightly packed side-chains and obtain complete description of folding from the random coil through molten globule to the native state.